I called the AVT support this morning and was told, that the Guppy camera doesn't correctly initialize the register that keeps the type of the used bayer pattern. I created a ticket, with a new patch, targeting this camera in particular. https://code.ros.org/trac/ros-pkg/ticket/5063 Sebastian Am 13.07.2011 15:11, schrieb Jack O'Quin: > On Wed, Jul 13, 2011 at 6:18 AM, Sebastian Schneider wrote: >> Hello, >> >> I tried to obtain images from an AVT Guppy using the nodelet configuration, >> as shown below: >> >> >> >> > args="manager" /> >> >> >> > args="load camera1394/driver camera_nodelet_manager"> >> >> >> >> >> >> >> >> > args="load image_proc/debayer camera_nodelet_manager"> >> >> >> >> >> >> >> >> >> >> >> >> However, the bayer pattern reconstruction is using the pattern rggb, instead >> of bggr, as it was told. Changing the pattern using the dynamic_reconfigure >> GUI, also does have no effect. >> >> I tried to trace the problem down and I found out, that as my camera is set >> to work in the format7 video-mode, thus most of the work is happening in >> camera_drivers/camera1394/src/nodes/format7.{h,cpp}. I looked into the files >> and found out, that the desired pattern is no where set in the code. >> Instead, regardless of the configured value, the pattern type is always read >> from the camera, using the dc1394_format7_get_color_filter function. The >> corresponding function to set the value was removed from libdc1394 a long >> time ago (God knows why). >> >> Since I don't know how to manipulate this value, I wrote a patch for >> format7.{h,cpp} that basically stores the value of the desired bayer >> pattern, and passes this value on in the corresponding field of >> sensor_msgs::Image in the function unpackData. The patch is appended to this >> email. >> >> I'm not sure if what I do there is right, even though I now have color >> images with the correct color (the sky is really blue now). Maybe someone >> responsible for the developement of this driver can have a look at it. > > OK. Please open a bug ticket for this problem. > > As best I can recall, the driver assumes that the Format7 Bayer > pattern reported by the device is correct and overrides user input in > that case. Maybe that's not true for your device? > > Thanks for the report,